Data Monitoring in Clinical Trials David L DeMets Curt D Furberg Lawrence M Friedman Editors Data Monitoring in Clinical Trials A Case Studies Approach With 40 Illustrations David L DeMets Department of Biostatistics and Medical Informatics University of Wisconsin Medical School Madison, WI 53972-4675 USA Curt D Furberg Department of Public Health Sciences Wake Forest University School of Medicine Winston-Salem, NC 27157 USA Lawrence M Friedman Bethesda, MD USA Library of Congress Control Number: ISBN-10: 0-387-20330-3 ISBN-13: 978-0387-20330-0 Printed on acid-free paper © 2006 Springer Science+Business Media, Inc All rights reserved This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, Inc., 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights Printed in the United States of America springeronline.com (MP) Preface Monitoring of clinical trials for early evidence of benefit and harm has gotten considerable attention.1 More formal guidelines and requirements2–4 have evolved in recent years, but in fact monitoring of trials is a practice that has been going on for almost four decades.5 For trials that involved conditions or interventions with serious risks, such as mortality or major morbidity, the tradition and policy has been to have an independent monitoring committee to review accumulating data for evidence of harm or convincing benefit that would require modifying or terminating a trial early During the past four decades, many trials have had monitoring committees to assume this responsibility.With the new emphasis on monitoring, this type of activity is increasing dramatically as the number of clinical trials being conducted to evaluate new interventions for patients or participants with serious risk or serious outcomes also increases.For example,policies of the National Institutes of Health (NIH) in the United States (US) call for monitoring committees for all phase III trials.2 Guidelines of the US Food and Drug Administration suggest such committees for trials of high-risk interventions or patients at high risk.3 As the number of monitoring committees increases, the challenge exists to pass along the experiences and best practices of the monitoring process to colleagues who are assuming this responsibility for the first time Textbooks such as the one by Ellenberg, Fleming, and DeMets6 provide many of the basic principles for monitoring committees Other texts such as those by Friedman, Furberg, and DeMets;7 Meinert;8 Pocock;9 Jennison and Turnbull;10 and Piantidosi11 provide statistical fundamentals and methods for the design, monitoring, and analysis of clinical trials This text is intended to complement those texts by providing a collection of examples or case studies of monitoring experiences from a variety of trials across different disease disciplines Each case study will describe the background of the individual trial, summarize the overall results, review the critical issues that emerged in the monitoring of the trial, and finally reflect on the lessons learned from that trial All of the examples presented share the complexity of the process of monitoring and the lesson that no single rule or algorithm can replace the wisdom and judgment of a monitoring committee.Through these examples, we hope to share the experience of these past committees and pass along some of their sometimes hard-earned wisdom Selection of the case studies was largely based on the collective experiences of the editors and their interactions with colleagues involved with clinv vi Preface ical trials Many of the 29 examples are from the field of cardiology, where the practice of monitoring committees was established early However, there are examples from other disciplines Regardless of the disease, many of the lessons learned and practices are useful for any trial Individual colleagues were invited to present the monitoring experience of a trial they were involved with as they saw it and experienced it.Their presentations and discussions not necessarily represent the official view of either the trial sponsor, the trial investigators, or the trial monitoring committee We have tried to get representation from each of these constituencies on many of the trials when possible For most of the past four decades, the existence and practice of monitoring committees has not been widely recognized or understood Our belief is that clinical research will benefit with better understanding of the process by both the research community and the interested public The intended audience for this book are those who are planning to serve on a monitoring committee or are already on one and wish to gain further insight into the monitoring and decision-making process We also believe that these examples will be useful to investigators as they design their trials and propose monitoring procedures; to sponsors, who typically receive monitoring committee recommendations, and to regulatory agencies, who often must review the results of trials that have been monitored by a committee In addition, Institutional Review Boards may benefit from these case studies since they ultimately have responsibility for protecting participants at the local level but must rely on the monitoring committee process for most multicenter trials and increasingly for institutional trials Journal editors, sciences writers, and practicing physicians may also find these case studies instructive Over the past four decades, many individuals have served on monitoring committees and participated in the monitoring of many challenging studies We wish to thank all of those individuals who have contributed directly or indirectly to the practice of monitoring and from whose experience we all have benefitted We have listed in Appendix the individuals who have served on the committees for the trials presented as case studies in this book and wish to thank them in particular ACKNOWLEDGMENTS We also want to thank the many contributors to the drafting of these case studies.We have listed them in the section which follows.They contributed their experiences because of their commitment to clinical trials, the monitoring process, and to teaching the next generation of clinical trial researchers about the important process of monitoring trials for early evi- Preface vii dence of benefit or harm We are grateful that they accepted our invitation and persevered through the drafts and editing process We would also like to acknowledge the substantial contributions by Ms Suzanne Parman for her editorial and logistical support.Without her dedication this text could not have been completed in a timely fashion David L DeMets Curt D Furberg Lawrence M Friedman REFERENCES Shalala D: Protecting research subjects–what must be done 2000 N Engl J Med 343: 808–810 National Institutes of Health 2000 Further Guidance on a Data and Safety Monitoring for Phase I and Phase II Trials, NIH Guide, June 5, 2000 http://grants.nih.gov/grants/guide/ notice-files/NOT-OD-00-038.html US Food and Drug Administration 2001 Draft Guidance for Clinical Trial sponsors on the establishment and operation of Clinical Trial Data Monitoring Committees Rockville, MD: FDA http://www.fda.gov/cber/gdlns/clindatmon.htm Food and Drug Administration, Department of Health and Human Services 1998 International Conference on Harmonisation: Guidance on statistical principles for clinical trials; availability Federal Register Vol 63, No 179:49583–49598 Greenberg Report: Organization, review, and administration of cooperative studies 1988 Control Clin Trials 9:137–148 Ellenberg S, Fleming T, DeMets D 2002 Data Monitoring Committees in Clinical Trials: A Practical Perspective John Wiley & Sons, Ltd.,West Sussex, England Friedman LM, Furberg CD, DeMets DL 1998 Fundamentals of Clinical Trials.Third Edition, Springer-Verlag, New York Meinert CL 1986 Clinical Trials: Design, Conduct, and Analysis Oxford University Press, New York Pocock S 1983 Clinical Trials: A Practical Approach John Wiley & Sons, Ltd.,West Sussex, England 10 Jennison C,Turnbull BW 1999 Group Sequential Methods With Applications to Clinical Trials Chapman and Hall/CRC, Boca Raton and London 11 Piantadosi S 1997 Clinical Trials: A Methodologic Perspective John Wiley & Sons, Inc., New York Contributors Susan Anderson Department of Biostatistics and Medical Informatics, University of Wisconsin, Madison,Wisconsin Data Monitoring in the Prospective Randomized Milrinone Survival Evaluation: Dealing with an Agonizing Trend Alex Bajamonde Genentech Inc., San Francisco, California Making Independence Work: Monitoring the Bevacizumab Colorectal Cancer Clinical Trial Jean-Pierre Boissel Clinical Pharmacology Department, Claude Bernard University, Lyon, France Stopping the Randomized Aldactone Evaluation Study Early for Efficacy Byron W Brown, Jr Stanford, California The Nocturnal Oxygen Therapy Trial Data Monitoring Experience: Problem with Reporting Lags Julie Buring Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School; Boston, Massachusetts Stopping the Carotene and Retinol Efficacy Trial: The Viewpoint of the Safety and Endpoint Monitoring Committee Paul L Canner Maryland Medical Research Institute, Baltimore, Maryland Breaking New Ground: Data Monitoring in the Coronary Drug Project Heidi Christ-Schmidt Statistics Collaborative,Washington, D.C Making Independence Work: Monitoring the Bevacizumab Colorectal Cancer Clinical Trial ix x Contributors Charles Clark Departments of Medicine, Pharmacology and Toxicology, School of Medicine, Indiana University, Bloomington, Indiana Early Termination of the Diabetes Control and Complications Trial Patricia Cleary The Biostatistics Center, The George Washington University, Rockville, Maryland Early Termination of the Diabetes Control and Complications Trial Robert Cody Department of Internal Medicine, Division of Cardiology, University of Michigan,Ann Arbor, Michigan Data Monitoring in the Prospective Randomized Milrinone Survival Evaluation: Dealing with an Agonizing Trend Theodore Colton Department of Epidemiology, Boston University School of Public Health, Boston, Massachusetts Challenges in Monitoring the Breast Cancer Prevention Trial Joseph P Costantino Department of Biostatistics, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania Challenges in Monitoring the Breast Cancer Prevention Trial Oscar Crofford Department of Medicine,Vanderbilt University, Nashville,Tennessee Early Termination of the Diabetes Control and Complications Trial Jeffrey A Cutler National Heart, Lung, and Blood Institute, Division of Epidemiology and Clinical Applications, National Institutes of Health, Bethesda, Maryland Data Monitoring in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial: Early Termination of the Doxazosin Treatment Arm Barry R Davis The University of Texas Health Science Center at Houston, School of Public Health, Houston,Texas Data Monitoring in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial: Early Termination of the Doxazosin Treatment Arm Contributors xi David L DeMets Department of Biostatistics and Medical Informatics, University of Wisconsin, Madison,Wisconsin Data and Safety Monitoring in the Beta-Blocker Heart Attack Trial: Early Experience in Formal Monitoring Methods Data Monitoring for the Aspirin Component of the Physicians’ Health Study: Issues in Early Termination for a Major Secondary Endpoint The Data Monitoring Experience in the Cardiac Arrhythmia Suppression Trial: The Need To Be Prepared Early The Nocturnal Oxygen Therapy Trial Data Monitoring Experience: Problem with Reporting Lags Kenneth Dickstein Cardiology Division, Stavanger University Hospital, Stavanger, Norway Data Monitoring Experience in the Moxonidine Congestive Heart Failure Trial Fred Ederer Bethesda, Maryland Assessing Possible Late Treatment Effects Early: The Diabetic Retinopathy Study Experience Susan S Ellenberg University of Pennsylvania School of Medicine, Center for Clinical Epidemiology and Biostatistics, Philadelphia, Pennsylvania FDA and Clinical Trial Data Monitoring Committees Frederick Ferris Division of Epidemiology and Clinical Research, National Eye Institute, National Institutes of Health, Bethesda, Maryland Early Termination of the Diabetes Control and Complications Trial Jan Feyzi Department of Biostatistics and Medical Informatics, University of Wisconsin, Madison,Wisconsin Data Monitoring Experience in the Metoprolol CR/XL Randomized Intervention Trial in Chronic Heart Failure: Potentially High Risk Treatment in High Risk Patients Dianne M Finkelstein Biostatistics Center, Massachusetts General Hospital; Harvard Medical School, Boston, Massachusetts xii Contributors Data Monitoring Experience in the AIDS Clinical Trials Group Study #981: Conflicting Interim Results Norman Fost Departments of Pediatrics and Medical History and Bioethics, University of Wisconsin, Madison,Wisconsin Monitoring a Clinical Trial with Waiver of Informed Consent: Diaspirin Cross-Linked Hemoglobin for Emergency Treatment of Post-Traumatic Shock Gary Francis Department of Cardiology, Cleveland Clinic Foundation, Cleveland, Ohio Data Monitoring Experience in the Moxonidine Congestive Heart Failure Trial Lawrence M Friedman Bethesda, Maryland Data and Safety Monitoring in the Beta-Blocker Heart Attack Trial: Early Experience in Formal Monitoring Methods The Data Monitoring Experience in the Cardiac Arrhythmia Suppression Trial: The Need To Be Prepared Early Curt D Furberg Department of Public Health Sciences, Wake Forest University School of Medicine,Winston-Salem, North Carolina Stopping the Randomized Aldactone Evaluation Study Early for Efficacy Stopping a Trial for Futility:The Cooperative New Scandinavian Enalapril Survival Study II Trial Lessons from Warfarin Trials in Atrial Fibrillation: Missing the Window of Opportunity Saul Genuth Division of Clinical and Molecular Endocrinology, Department of Medicine, University Hospitals of Cleveland, Case Western Reserve University, Cleveland, Ohio Early Termination of the Diabetes Control and Complications Trial Stephen L George Department of Biostatistics and Bioinformatics, Director, Cancer Center Biostatistics, Duke University Medical Center, Durham, North Carolina Controversies in the Early Reporting of a Clinical Trial in Early Breast Cancer CASE 29 Making Independence Work: Monitoring the Bevacizumab Colorectal Cancer Clinical Trial Janet Wittes Eric Holmgren Heidi Christ-Schmidt Alex Bajamonde ABSTRACT Genentech’s phase III study of colorectal cancer aimed to show that administration of bevacizumab, its recombinant human anti-vascular endothelial growth factor antibody, to patients with metastatic colorectal cancer would reduce mortality The four-member DMC was responsible for reviewing real-time safety data, selecting the experimental arm of the study at an interim analysis, and assessing efficacy at a second interim analysis Genentech and the entire study team remained blinded to treatment allocation during the course of the trial.The study, which proceeded to its planned end, showed an estimated hazard ratio for death of 0.66 (p < 0.001) On the basis of this study, the FDA approved bevacizumab for patients with metastatic colon cancer This example shows how a DMC can make complicated decisions and recommendations even when neither it nor the statisticians reporting to it participate in any other way with the conduct of the study INTRODUCTION AND BACKGROUND The late 1990s saw the refinement of our biological understanding of angiogenesis, the development of blood vessels Because angiogenesis contributes to the growth of human cancers, scientists theorized that interfering with angiogenesis should retard tumor growth This theory led various drug companies to develop monoclonal anti-vascular endothelial growth factor antibodies as potential treatments for cancer One such antibody is Genentech’s bevacizumab [Avastin®], which showed promising effects 360 Making Independence Work 361 when administered with fluorouracil plus leucovorin in a phase II trial of colorectal cancer.1 Some early-phase studies with the product in other cancers had suggested serious toxicities associated with its use, in particular, excess major bleeding, thrombosis, hypertension, proteinuria, and severe diarrhea Thus, proceeding with a phase III study required careful monitoring of these safety events Genentech chose a four-member Data Monitoring Committee (DMC) and asked it to review safety data in essentially real time to ensure that the rate of serious adverse events in patients given bevacizumab was not unacceptably high When the study was being designed, the standard of care for the treatment of metastatic colon cancer had been a combination of two chemotherapeutic agents, fluorouracil and leucovorin (FL) Genentech had studied bevacizumab in colorectal cancer in a phase II trial with the FL regimen as background therapy In 2000, when Genentech was ready to initiate its phase III study to demonstrate the effect of bevacizumab on mortality in colorectal cancer, optimal care had changed to a regimen consisting of irinotecan, fluorouracil, and leucovorin (IFL).2 The designers of the phase III trial selected the IFL regimen as the control arm because its efficacy was superior to that of FL; however, the IFL regimen was considerably more toxic.The choice of experimental regimen then became problematic To start a large phase III3 trial comparing the IFL regimen plus bevacizumab to IFL alone was risky, for if the combination proved unacceptably toxic, the study would stop early without adequately testing the effect of bevacizumab.On the other hand, selecting FL plus bevacizumab as the experimental arm could produce ambiguous results If the study showed FL plus bevacizumab to beless effective than IFL, the oncologic community would not be able to distinguish between the inferiority of the FL regimen or the lack of efficacy of bevacizumab If the experimental arm demonstrated superiority over the IFL regimen, oncologists would not know whether combining IFL with bevacizumab would be even more effective PROTOCOL DESIGN The sponsor and the investigators designed a study in colorectal cancer that started with a control group given IFL and a bevacizumab placebo It had two experimental arms, one that combined the antibody with FL and one that combined it with IFL.The primary endpoint was mortality; the secondary endpoint was time to progression The study began, therefore, as a three-arm trial of patients with metastatic colorectal cancer.The two IFL arms were blind; however, because the schedule of administration of the FL arm differed from the schedule for the IFL alone arm, the FL plus bevacizumab arm could not be blinded 362 Data Monitoring in Clinical Trials:A Case Studies Approach Integral to the design of the study was the DMC, which was to select the experimental arm during the course of the study When approximately 100 patients had been randomized to each of the three treatment arms, the DMC would review all data on safety.The DMC would then select either the bevacizumab plus FL or the bevacizumab plus IFL as the experimental arm; it was to select the combination with IFL as long as that arm was “not unsafe.” Having made the decision, the DMC would recommend randomizing only to the ILF alone and to the selected experimental arm The total sample size would be approximately 900 patients—400 each in the control and selected experimental arm, 100 in the arm not selected After determination of the experimental arm, the study was to continue until 385 deaths had occurred unless the DMC recommended stopping the study at its single efficacy interim analysis The interim analysis for efficacy was to occur halfway through the trial, that is, when a total of 193 deaths had occurred in the control (IFL) and the selected treatment arm The DMC would use an O’Brien–Fleming3 spending function to assess whether to stop for efficacy The study design specified a log-rank test to compare mortality in the two groups with a two-sided type error rate of 0.05.This design had approximately 80% power to detect a hazard ratio of 0.75 for death in the group given the experimental treatment compared to the control group DATA MONITORING EXPERIENCE Genentech established a Data Monitoring Committee composed of three medical oncologists, all of whom treated patients with colorectal cancer and were involved in clinical trials of the disease In addition, the DMC had one statistician.The DMC reviewed concurrently this trial and a phase II trial in patients with colorectal cancer who were too fragile to be given the IFL regimen.This second trial was two-arm study of bevacizumab plus FL compared to FL alone.While the data from this second trial played no formal role in judgments about the phase III trial, the DMC used the data, especially the data on safety, in statistically informal ways to provide additional insight into the use of bevacizumab in colorectal cancer The statistician on the committee also was the statistician for a concurrent study of bevacizumab in breast cancer; again, while he did not use the data from the breast cancer study explicitly in reviewing the phase III trial in colorectal cancer, the information provide additional qualitative insights into the use of the product Statistics Collaborative, a statistical consulting firm with no other relationship to the three studies, presented data to the DMC for each study Therefore, like the statistician on the DMC, staff at Statistics Collaborative had access to information from all three ongoing studies Genentech and the Making Independence Work 363 investigators in all three studies were blind to treatment allocation (except for the FL and bevacizumab arm in the phase III study) at the individual patient level.Importantly,they remained blind to summary data by treatment, even for the FL and bevacizumab arm During the trial, the DMC had three distinct roles First, it was charged with monitoring safety In light of observations from early-phase studies, the DMC was to review data on specific adverse events every two weeks until it met to choose the experimental arm.At that point, the DMC would review safety data monthly Genentech designed four special case report forms, one for each of the following events: (a) thromboembolic events, (b) major bleeding, (c) severe hypertension, and (d) severe diarrhea.The DMC was to monitor these so-called “targeted events”as well as all serious adverse events Every two weeks, Genentech downloaded a database with information from these forms and Statistics Collaborative prepared a brief report for the DMC The DMC met by telephone to review the data and to decide whether to recommend continuation of the trial Second, the DMC was to choose the experimental arm on the basis of safety Genentech and the investigators did not provide the DMC with formal guidelines for this decision; rather, it was to use its collective judgment to assess whether the bevacizumab plus IFL regimen was “not unsafe.” Finally, it was to review efficacy The predefined statistical guideline for stopping for efficacy specified the DMC was to look formally at efficacy at the halfway point in the trial If at that time the p-value for benefit was 0.0018, the DMC could recommend stopping for efficacy The first patient was randomized in September 2000 Each in-person DMC meeting began with an open session with the sponsor to report on the status of the trial The representatives from the sponsor were staff directly involved in the design and conduct of the study At the closed session, attended only by the DMC and the statisticians reporting to it, the DMC reviewed the data At the end of the meeting, the DMC reported its recommendations by telephone and in writing to a committee at Genentech composed of clinical experts—a statistician and physicians— with no direct responsibility for the conduct of the study.This Data Review Board, which served as a buffer between the DMC and the study team, had the authority to review unblinded data if the DMC made a recommendation to stop the study early The study had three such meetings The first, or introductory meeting, dealt with the design of the study, the hypothesized action of bevacizumab, and the DMC’s charter At the second in-person meeting, the DMC met to select the experimental arm.The third meeting dealt with the formal interim analysis for safety In addition, the board met monthly by telephone with Statistics Collaborative to discuss safety data 364 Data Monitoring in Clinical Trials:A Case Studies Approach The real-time reporting of the data made the process complex Data came from Genentech to Statistics Collaborative in several different streams and structures Because Statistics Collaborative was not involved in collecting the data and Genentech was blind to treatment allocation, the two groups met by telephone weekly to discuss procedures and the data themselves For the interim in-person meetings Genentech provided Statistics Collaborative with all the data from the study, not just the data planned for the DMC This procedure, although quite labor intensive, allowed Statistics Collaborative to understand the nature of the data adequately enough to report accurately to the DMC Early Real-Time Safety Data Almost from the beginning of the study, the serious and targeted adverse event rates were higher in the IFL arms than in the bevacizuman plus FL; however, because the rates in the bevacizumab plus IFL arm were very similar to the rates in the placebo plus IFL arm, the DMC did not recommend any change in protocol or issue any statement about adverse events The biweekly reports changed in character over time At first, the reports used graphs we all dubbed “LJ Pictures,” after LJ Wei, the statistician on the DMC These pictures displayed, on a patient-by-patient basis, the time course of all serious and targeted adverse events.As the weeks progressed and the dataset grew, the format of the reports changed to summarize more succinctly the patterns of events The 300-Patient Meeting: Choosing the Treatment Arm At its 300-patient in-person meeting, the DMC reviewed data on safety for the three arms.Although patients in the bevacizumab plus IFL arm had somewhat higher adverse event rates than did those in the bevacizumab plus LF arm (Table 1), the DMC judged the arm as “not unsafe.” Consistent with its charge, it recommended proceeding with the bevacizumab plus IFL arm and not randomizing further to the bevacizumab plus LF arm It did make two recommendations related to safety for both IFL arms It recommended that the General Medical Concerns section of the exclusion criteria indicate that patients older than 65 years are at increased risk of irinotecanassociated diarrhea It also recommended that the study exclude patients whose total bilirubin exceeded 1.6 mg/dl because other studies had shown that patients with Gilbert’s syndrome, who have elevated bilirubin measurements, are at higher risk of adverse events when they are on the IFL regimen While its charter had specified that real-time safety monitoring would end after the first in-person meeting, the DMC opted to continue its real-time monitoring monthly because it felt that 100 patients in the bevacizumab plus IFL arm constituted an insufficient safety database Making Independence Work 365 Table Estimated Percentage of People Experiencing Specific Adverse Events Within the First Five Months of Therapy People with at least one— Serious or targeted adverse event Serious adverse event Grade or bleeding Grade or diarrhea Thromboembolic event Episode of serious hypertension Grade or proteinuria* Experimental arm 1: bevacizumab plus IFL (n = 113) Experimental arm 2: bevacizumab plus FL (n = 112) Control arm: IFL (n = 110) 45 35 29 27 23 25 15 12 20 20 0 0 * Note: Grades and refer to serious adverse events and life-threatening adverse events, respectively The Efficacy Analysis At its third in-person meeting, the DMC reviewed data from the two enrolling arms to judge whether to recommend stopping the study on the basis of efficacy of the bevacizumab plus IFL arm.The estimated hazard ratio for mortality was 0.64 The DMC, impressed with the large reduction in mortality, considered recommending stopping for efficacy, but decided that because the data had not crossed the prespecified boundary for efficacy (observed p = 0.003; critical p = 0.0018), a recommendation for early stopping would jeopardize the study because the results would not be statistically convincing The DMC also reviewed safety data of all patients enrolled in the study, including the bevacizumab plus FL arm that was closed to enrollment.This review was the last safety review for this study Ending the Study The last patient entered in May 2002 The study ended as planned The median duration of survival was 20.3 months in the IFL plus bevacizumab group and 15.6 months in the IFL plus placebo group for an estimated hazard ration of 0.66 for death (p < 0.001).4 In 2004, the FDA granted approval for bevacizumab in the treatment of metastatic colorectal cancer.5 LESSONS LEARNED This happy case history may appear boring; it may remind the reader of Tolstoy’s introduction to Anna Karenina:“Happy families are all alike; every unhappy family is unhappy in its own way.” The retelling sounds as if the 366 Data Monitoring in Clinical Trials:A Case Studies Approach study, from the vantage point of the DMC, proceeded smoothly.The DMC had three responsibilities, all of which it discharged responsibly.The antibody was highly effective so that at the end of the study everyone—sponsor, investigators, the DMC, the patients, and the oncology community at large—benefitted from the successful implementation of the trial But the calm, orderly history masks difficulties that occurred with the DMC during the study Managing real-time data from a study with many centers poses huge logistical challenges for the sponsor and for the reporting statistician The successful monitoring of a trial of this type requires that the sponsor and the reporting statistician develop mutual trust Because the patients in this trial came from 164 clinical sites, the logistics of implementing this study required considerable effort on the part of the sponsor The sponsor necessarily focused on the implementation of the study.To pretend that all interactions among the two pairs of authors of this paper were smooth during the years of the study would be to ignore the inevitable tensions that arise when two groups are working together with complementary but very different roles Because Genentech remained blind to the effects of treatment, it often found Statistics Collaborative’s requests, which echoed the questions from the DMC, unreasonable For its part, Statistics Collaborative found Genentech’s reluctance to comply with requests frustrating.The basic problem we both faced was that Statistics Collaborative could not tell Genentech the reason for its requests and Genentech could not tell Statistics Collaborative about some of the constraints under which it operated The two groups had to develop enough trust to allow us to proceed as each party thought best even though we were ignorant of the reasons for the other’s needs.At one point, two of us (J.W and A.B.) took a long walk along a Pacific beach knowing that the calm lapping of the waters would lead us to better cooperation during the course of the study In trials of toxic therapy, a DMC with direct experience treating patients of the type being studied provides invaluable insight into safety In many disease areas, a DMC can look at safety infrequently Because of the toxicity of chemotherapy, trials in cancer require frequent monitoring of safety Part of the success of this DMC was the expertise of the clinicians on the DMC and their direct experience with patients of the type under study They were able to assess the importance of the particular events that were occurring in the context of knowledge of other chemotherapeutic regimens A statistician experienced in DMCs and intellectually engaged in the study can contribute greatly to the process of safety monitoring The statistician on this DMC played a crucial role He pushed for various ways of looking at the data Of particular interest to him, and ultimately to rest of the DMC, was the time course of adverse events Making Independence Work 367 The reporting structure for a DMC, in bypassing the study team, can afford an extra level of protection of the integrity of the study Independent data monitoring of a complex protocol requires considerable work and expense In retrospect, everyone involved in the DMC for this study believes the charge to, and the actions of, the DMC strengthened the study In particular, the ability to proceed without having selected the final experimental arm saved many months over alternative designs Only the use of a DMC, and the separation of it from the operation of the study, could have allowed this selection during the course of a study Nonetheless, this study showed all of us involved how time-consuming and labor-intensive—and hence expensive— this type of activity can be Before embarking on such a large undertaking, the sponsor should satisfy itself of the necessity of such a structure ROLES AND ACKNOWLEDGMENTS The authors of this chapter included two statisticians at Genentech, the study statistician (E.H.) and the head of oncology statistics (A.B.) as well as the two statisticians at Statistics Collaborative (H.C.S and J.W.) who presented the data to the DMC.The DMC was composed of three oncologists— Richard Schilsky, M D (chair), Robert Mayer, M.D.; Alan Venook, M.D.—and one statistician (Lee-Jen Wei, Ph.D.) Herbert Hurwitz, M.D., was the principal investigator, and William Novotny, M.D., was the medical monitor REFERENCES Kabbinavar F, Hurwitz HI, Fehrenbacher L, Meropol NJ, Novotny WF, Lieberman G, et al 2003 Phase II, randomized trial comparing bevacizumab plus fluorouacil (FU)/leucovorin (LV) with FU/LV alone in patientts with metastatic colorectal cancer J Clin Oncol 21:60–65 Saltz LB, Cox JV, Blanke C, Rosen LS, Fehrenbacher L, Moore MJ, et al 2000 Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer N Engl J Med 343:905–914 O’Brien P, Fleming T 1979.A multiple testing procedure for clinical trials Biometrics 35:549–556 Hurwitz H, Fehrenbacher L, Novotny W, Cartwright T, Hainsworth J, Heim W, et al 2004 Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer N Engl J Med 350:2335–2342 Genentech, Inc Label for AvastinTM (Bevacizumab) U.S BLA Amendment: Bevacizumab— Genentech, Inc Februrary 2004, 27pp http://www.fda.gov/cder/foi/label/2004/ 125085lbl.pdf APPENDIX Data Monitoring Committee Members The following individuals have served on one or more monitoring committees Abeloff, Martin Applegate,William Armstrong, Paul W Ayers, Stephen M Ballintine, Elmer Bearman, Jacob, E Berge, Kenneth Berry, Donald A Blackburn, Henry Blum, Ronald H Boissel, Jean-Pierre Boyle, Edwin Bristow, J David Brown, B.William Bruce, Robert Buring, Julie Cairns, John Califf, Robert Campbell, Ronnie Canner, Paul L Carleton, Richard Carter, Michele Cassel, Cristine Chalmers,Thomas Chatterjee, Kashinath Childress, James F Clark, Charles Clements, Rex S Jr Cody, Robert J 368 Cohen, Lawrence M Collins, Rory Colton,Theodore Cooper, Edward Cornfield, Jerome Cowles,Andrus Crofford, Oscar Cryer, Henry M III Cutter, Gary Davis, C Edward Davis, Matthew DeMets, David L Deykin, Daniel Dickstein, Kenneth Ederer, Fred Erban, John Kalil Feigl, Polly Ferdinand, Keith Ferris, Frederick Fisch, Charles Fisher, Lloyd Fisher, Marian Fleming,Thomas R Ford, Ian Fost, Norman Francis, Gary Freedman, Laurence Friedewald,William T Friedman, Ephraim Monitoring Committee Members 369 Friedman, Lawrence M Furberg, Curt D Gent, Michael Gifford, Raymond Gillette, James Goldstein, Sidney Grant, Igor Hainline,Adrian Halperin, Max Hamilton, Michael P Harrington, David Harris, Jay R Hart, Robert Hennekens, Charles H Higgins, Millicent Hillis,Argye Hirsh, Jack Hochman, Judith Hulka, Barbara Hutchison, George B Iber, Frank Johnson, Susan Johnstone, David Judd, Howard Julian, Desmond G Kimbel, Philip Klatskin, Gerald Kligfield, Paul D Klimt, Christian R Knatterud, Genell Krome, Ronald Kulbertus, Henri Lachin, John Lasagna, Louis Lebovitz, Harold Lee, Stephanie J Levy, Robert Lewis, Roger J Lilienfeld,Abraham Malchon, Jean K Massie, Barrie Mayer, Kenneth H Mayer, Robert Meinert, Curtis L Menkes, Harold Miller,Anthony Miller, Dayton T Miller, Max Miller,William F Naughton, John Newman, Elliot Nies,Alan Norton, Edward Packer, Milton Palmberg, Paul Patz,Arnall Pitt, Bertram Pocock, Stuart Pollard, James Pollard, Richard B Pratt, Lois Prineas, Ronald Prout,Thaddeus Rahal, James J Rapaport, Elliot Rosner, Bernard Ruskin, Jeremy Ryan,Thomas Sackett, David Sackner, Marvin A Saudek, Christopher Schiffrin,Alicia Schilsky, Richard Sears, Marvin Sheps, Sheldon Smith,William M Snapinn, Steven Sobel, Burton Stamler, Jeremiah Stone, Neil Strauss, Harold Thompson, Simon Tilley, Barbara Turner, Robert 370 Data Monitoring in Clinical Trials:A Case Studies Approach Vachon, Louis Venook,Alan Walters, LeRoy Wang, Dualo Washington, Geraldine Wedel, Hans Wei, Lee-Jen Whitley, Richard J Wilhelmsen, Lars Wilkens, Robert Williams, O Dale Wittes, Janet Wyse, D.G Zukel,William APPENDIX Case Study Acronym Key (Title) ACTG: ALLHAT: AIDS Clinical Trials Group Study #981 Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial BCPT: Breast Cancer Prevention Trial BHAT: Beta-Blocker Heart Attack Trial CALGB: Cancer and Leukemia Group B CAPRICORN: Carvedilol Post-Infarct Survival Control in Left Ventricular Dysfunction Study CARET: Carotene and Retinol Efficacy Trial CAST: Cardiac Arrhythmia Suppression Trial CDP: Coronary Drug Project CHARM: Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity CONSENSUS II: Cooperative New Scandinavian Enalapril Survival Study II CURE: Clopidogrel in Unstable Angina To Prevent Recurrent Ischemic Events Trials DCCT: Diabetes Control and Complications Trial DCLHB Diaspirin Cross-Linked Hemoglobin Trial DRS: Diabetic Retinopathy Study HERS: Heart and Estrogen/progestin Replacement Trial HOPE: Heart Outcomes Prevention Evaluation MERIT-HF: Metoprolol CR/XL Randomized Intervention Trial in Chronic Heart Failure MOXCON: Moxonidine Congestive Heart Failure Trial NOTT: Nocturnal Oxygen Therapy Trial PHS: Physicians Health Study PROMISE: Prospective Randomized Milrinone Survival Evaluation RALES: Randomized Aldactone Evaluation Study RESOLVD: Randomized Evaluation of Strategies for Left Ventricular Dysfunction Pilot Study SPAF I: Stroke Prevention in Atrial Fibrillation I Trial TOXO: AIDS Toxoplasmic Encephalitis Study INDEX A A posteriori hypothesis, 188 A priori hypothesis, 188 ACTG, 22, 53, 109–117 ACTG-didanosine trial, 41 Acquired immunodeficiency syndrome (AIDS), 109–117 Agonzing negative trend, 209, 225, 242, 262 AIDS Clinical Trials Group (ACTG), 109–117 AIDS trials, 109–117, 320 Aldactone (see RALES) ALLHAT, 15, 17, 20, 179, 248–259 Aspirin trial, 73–83 Asymmetric boundaries, 143, 169, 198 Asymmetric monitoring guidelines, 21 Atrial fibrillation trial, 282, 312 B BAATAF, 316 Bayesian monitoring, 133 BCPT, 21, 28, 118–135 Benefit vs risk, 21 Beta-Blocker Heart Attack Trial (see BHAT) Beta carotene studies, 220–226 Betaseron, 45–47 Bevacizumab, 360–367 BHAT, 32, 53, 64–72 Boston Area Anticoagulation Trial for Atrial Fibrillation (see BAATAF) Breast cancer trial (see BCPT), 118, 282, 346 C Canadian Atrial Fibrillation Anticoagulation (CAFA) Study, 316 CALGB, 347 Cancer and Leukemia Group B (see CALGB) Cancer prevention trials, 220 Cancer trials, 269, 282, 346, 360 Candesartan (see CHARM, RESOLVD) CAPRICORN, 15, 16, 28, 33, 281, 337–345 CARET, 15, 17, 32, 179, 180, 220–227 Cardiac arrhythmia suppression trial (see CAST) Cardiac arrhythmia trials, 198 Cardiovascular trials, 64, 73, 136, 148, 158, 166, 183, 198, 209, 236, 248, 260, 302, 330, 337 Carotene (see CARET) Carvedilol (see CAPRICORN) 372 CAST, 18, 20, 24, 30, 180, 198–208 CDP, 4, 15, 17, 20, 29, 34, 179, 180, 183–197, 296 CHARM, 26, 28, 53, 166–175 CHESS trial, 43 Chlorthalidone, 179 Clofibrate, 192–194 Clopidogrel (see CURE) Collaborative antiviral study group, 286 Colorectal cancer trial, 282, 360 Complex decision process, 15, 81, 98, 116, 183, 198, 209, 220, 232, 240, 292, 302, 337, 360 Composite outcomes, 26–28 Conditional power, 68, 76, 191, 215, 253, 258 Conflict of interest, Conflicting outcomes, 22, 112, 131, 146, 161, 236 CONSENSUS II, 15, 18, 283, 302–311 CONVINCE, 35, 36 Copenhagen AFASAK study, 313 Coronary Drug Project (see CDP) CURE (Clopidogrel), 22, 54, 158–165 D Daclizumab, 269–277 Data flow, 71, 154, 198, 203, 275, 292, 299, 326 Data issues, 23–25 Database access for DMC, 230, 364 DCCT, 17, 31, 33, 53, 93–108 Degree of needed evidence, 20 Declaration of Helsinki, 20 Dextrothyroxine, 188–191 Diabetes trials, 53, 93 Diabetic Retinopathy Study (see DRS) Diaspirin, 228–235 Didanosine, 41, 42 DMC (see Monitoring committee) Doxazosin (see ALLHAT) DRS, 53, 55–63 DSMB (see Monitoring committee) DSMC (see Monitoring committee) E Early reporting of results, 60, 346 Early termination for benefit, 55, 64, 73, 85, 93, 109, 118, 136, 148, 158, 166 Index 373 for futility, 183, 302, 320 for harm, 183, 198, 209, 220, 228, 236, 248, 260 other, 34–36 Early trends, 18, 20, 25, 31 Early vs late benefits/risks, 55, 62, 106, 116, 128, 134, 164, 170 Eastern Cooperative Oncology Group (ECOG), 180, 347 Emergency research, 228–233 Enalapril (see CONSENSUS, RESOLVD) Encainide, 198–208 Encephalitis, 320 Endpoint adjudication, 77, 252, 266 Estrogen, 186–188, 191, 192 Estrogen/progestin, 236–247 Ethical issues, 19, 22 Ethical standards for monitoring, External data, 67, 88, 90, 136, 138, 220, 337, impact, 10 External information, 31, 33 Evolving trial relevance, 282, 320 F Flecainide, 198–208 Follow-up, post-trial, 33, 270–272 FDA monitoring committees, 40 Food and Drug Administration (FDA), 39 G Global index of risk/benefit, 123–128 Graft vs host disease, 269–276 Greenberg report, 4, H HA-IA, 42–44 Heart and Estrogen/Progestin Replacement Trial (see HERS) Heart Failure trial, 136, 148, 209, 260, 330, 337 Heart Outcomes Prevention Evaluation (see HOPE) Herpes simplex trial, 15, 282, 285–291 HERS, 21, 26, 30, 34, 180, 236–247 HIV/AIDS, 41, 42 Hormone therapy trials, 236 HOPE, 27, 54, 158–165 Hypertension trials, 248 I Imaginary scenarios, 88, 115, 125, 358 Informed consent, 228 waiver, 19, 228 Independent statistician, 140, 310, 328, 360, 366 Interferon trial in multiple sclerosis, 45–47 Interim analysis, unscheduled, 231, 272, 292, 352 Interim data sharing, 46, 47 Interim protocol change, 44 International Conference on Harmonization, 40 L Lagging death flow, 19 LIFE, 27 Lipid lowering trials, 183, 248 M MERIT-HF, 53, 136–147 Metoprolol Trial (see MERIT) Milrinone (see PROMISE) Monitoring external evidence, 31, 33, 222, 312, 218, 339, 344 guidelines, 10 overall event rates, 82, 340 rules vs guidelines, 64, 99, 121, 123, 139, 155, 164, 169, 193, 196, 207, 182, 257, 261, 268, 355, 363 secondary outcomes, 73, 81, 159, 241, 246 trial portfolio, 166, 276 Monitoring committee blinding, 107, 111, 156, 195, 203, 226, 230, 273, 337 charter, 15, 169, 206, 257, 332, 335, 350 communication, 11, 156, 195, 332, 335, 350 composition, 14 database access, 230, 364 ethical standards, function, independence, 282, 360, 367 lack of consensus, 17 leadership, 184, 282, 285 meetings, 6, 25 membership, post-trial responsibility, 268 prior experience, 106, 185, 207, 366 publication responsibility, 23 responsibilities, 14, 15 Moricizine, 198–208 MOXCON, 18, 21, 30 179, 260–268 Moxonidine (see MOXCON) Multiple outcomes, 6, 93, 118, 189, 195, 254, 293, 340 Multiple sclerosis interferon trial, 45 Myocardial infarction trials, 64, 183 N Negative trends, 26 NETT, 29 Niacin, 194 NIH-industry DMC model, 211, 217 Nocturnal Oxygen Therapy Trial (see NOTT) North Central Cancer Treatment (NCCT), 347 NOTT, 19, 24,281, 292–301 374 Index O One vs two tailed hypothesis, 200, 205, 271 Outcomes, Composite, 26 Conflicting, 22, 112, 131, 146, 161, 200, 202, 236 Primary vs secondary, 22 Surrogate, 30 P Pacilitaxel, 346–359 Physicians Health Study (see PHS) PHS, 22, 54, 73–84 Placebo control, 289 Policy board, 4, 57, 93, 107, 185 Post trial surveillence, 55, 195, 244 PRAISE, 23, 29 Primary vs secondary outcomes, 22 PROMISE, 30, 179, 209–219 Propranolol, 64–71 Prospective Randomized Milrinone Survival Evaluation (see PROMISE) Protocol modification, 15, 60, 111, 281, 337, 340, 342 PROVE-IT, 27 Publication of negative trials, 267 Q Quality of life assessment, 93, 139, 295 R RALES, 19, 23, 54, 148–157 Realtime safety monitoring, 143, 168, 364 Regulatory guidance, 39 Repeated analysis issues, RESOLVD, 15, 283, 330–336 Response to DMC recommendations, 224, 254, 265, 283, 330, 334 S Second opinion committees, 82, 224, 254, 282, 330, 334 Sepsis trial, 42–44 Sequential boundaries, 68, 98, 111, 121, 139, 151, 160, 169, 192, 202, 212, 215, 221, 251, 240, 271 Sharing interim data, 46, 47 Southwest Oncology Group (SWOG), 347 SPAF, 20, 53, 85–92, 315 Sponsor unblinding, 42 Statistical monitoring guidelines, 15, 64, 71, 93, 192–93, 202, 212, 240, 251, 305, 350 Stroke Prevention in Atrial Fibrillation Trial (see SPAF) Stroke trials, 85 Study chair unblinding, 170, 217, 326, 327 Subgroup issues, 28–31, 87, 281, 292, 298–99, 287–90, 217, Surrogate outcomes, 30 T Taxol®, 346–359 Termination for futility, 195, 244, 283, 302, 324 early, procedures, TOXO, 281, 320–329 Toxoplasmic encephalitis (see TOXO) T-PA trial in stroke, 44 Trial extension of follow-up, 196, 244 U Unblinding sponsor, 42 study chair, 170, 217, 326, 327 Unexpected large benefit, 88, 105 Unexpected negative trials, 203, 209, 230, 236, 242, 255, 257, 262, 271 Unscheduled interim analysis, 231, 272, 292, 352 Use of placebo control, 289 V Veterans Affairs Stroke Prevention in Nonthrematic Atrial Fibrillation Study (SPINAF), 317 W Warfarin trials, 312–319 Women’s Health Initiative, 34 Window of opportunity, 283, 312 ... Siebert for the DCCT Research Group Data Monitoring in the AIDS Clinical Trials Group Study #981: Conflicting Interim Results Dianne M Finkelstein Challenges in Monitoring the Breast Cancer Prevention... outcome data across the entire study When 12 Data Monitoring in Clinical Trials: A Case Studies Approach initially reviewing trial protocols, the IRBs should be informed about the plans for monitoring, ... by the monitoring committee, and the investigators and no statistical monitoring boundaries were pre-specified During the course of the trial, it became 16 Data Monitoring in Clinical Trials: A